Applications from biology and electronic commerce
often feature categorical variables with an enormous
number of levels and a power law (Zipf-like) distributions.
This talk addresses how to visualize the joint behavior
of two such categorical variables. Examples include
in-degree versus out-degree in large graphs, and the
identities of raters and rated items.  We use Zipf-Poisson
and Mandelbrot-Zipf-Poisson models to describe when
the sample data are representative of an underlying
process.